“The best-laid schemes o' mice an' men,” penned Robert Burns in his ode to mice. It is an astute observation of how our intelligence has outwitted theirs. Though I appreciate Burkart et al.'s attempt to synthesize a wild and wooly comparative literature on general intelligence in mice, men, and many other species, they have introduced a few red herrings and false premises that muddy the waters and undermine suggestions for future research.
The problem starts with the authors' initial premise: Scholars tend to view the mind as either domain-specific/modular (DSM) or domain-general (DG), and those who lean to DSM see the mind as predetermined and inflexible, and thus largely the work of nature. These views conflict with a theory of general intelligence. Burkart et al. claim that their framework shows “that human cognition involves elements of domain-specific and domain-general processes” (sect. 1.2.3, para. 4), and in contrast to prior views, “animal minds need not be bundles of specialized cognitive adaptations” (sect. 1.2.3, para. 4). But their premise is false as is their characterization of animal research. This commentary addresses these misconceptions and introduces some additional distinctions in order to productively explore how general intelligence evolved.
Those who have synthesized DSM perspectives (e.g., Pinker Reference Pinker1997) do not deny the existence or significance of DG capacities: evolutionarily ancient mechanisms that typically interface with and often constrain the outputs of each domain. Research on theory of mind (ToM), number, and language – domains often considered as modules – has long explored how executive functions interact with the computations and representations of each domain (Bradford et al. Reference Bradford, Jentzsch and Gomez2015; Soltész et al. Reference Soltész, Goswami, White and Szűcs2011). For example, delays in the expression of ToM and number competence are intimately related to the development of working memory, whereas performance on ToM tasks can be improved by lifting constraints that arise from inhibitory control or perseverative responses. Thus, although it is inaccurate to pigeonhole scholars as either DSM or DG, it is true that those who have explored the nature of DSM systems are more interested in them and in how they can be characterized on the basis of evolutionary theory. Similarly, although the generative computations that subserve language competence (but also other domains such as music, number, and ToM) have no limit, our capacity to produce or comprehend sentences is limited by working memory. Thus, although DSM-focused researchers tend to emphasize the nature of the representations and computations within a domain or module, they don't deny the existence or potentially constraining impact of DG processes.
Of relevance to the evolution of general intelligence is the underlying architecture of DSM systems. Here, too, Burkart et al. mischaracterize these as innate and inflexible. Research on faces reveals this error. Neurobiological studies in macaques and humans reveals dedicated circuitry that is consistent with a DSM perspective. However, this system matures slowly over time and depends on experience with faces as elegantly demonstrated by studies of individuals with early-appearing cataracts that were later removed (Rhodes et al. Reference Rhodes, Nishimura, de Heering, Jeffery and Maurer2017). A similar characterization applies to language, wherein there are core underlying computations and representations, some specific to language and others shared (Hauser & Watumull Reference Hauser and Watumull2016), but with experience selecting among the options to generate specific languages (e.g., French, English).
Lastly, it is simply not the case that nonhuman animals are perceived as mere bundles of modules, fixed and inflexible. Research on model systems such as aplysia and songbirds reveals both ancient, general mechanisms for learning and memory, as well as highly dedicated systems that nonetheless show plasticity. For example, although passerines acquire their song on the basis of specialized circuitry that enables vocal imitation, this same system requires specific input (e.g., species-specific song), is not engaged for other vocalizations (e.g., alarm calls), and in some species, shows plasticity throughout life as individuals create new songs each season. In addition, many researchers have recognized and detailed other DG processes that go beyond what Burkart et al. discuss. For example, there is considerable comparative work exploring the concept of “sameness,” analogical reasoning, and algebraic computations (Martinho & Kacelnik Reference Martinho and Kacelnik2016; Smirnova et al. Reference Smirnova, Zorina, Obozova and Wasserman2015; ten Cate Reference ten Cate2016). These are not part of the executive system, have not typically been linked to general intelligence, and yet they cut across domains and appear evolutionarily ancient.
Putting these strands together suggests that any approach to exploring the evolution of intelligence must consider the interaction between DSM and DG, understand the specificity of the content of DSM, examine a diversity of DG systems (i.e., beyond executive functions), and document how maturational changes in DG can impact the ontogeny of DSM. The content of a domain is particularly relevant as tasks within the general intelligence battery are often assumed to be part of a given domain without rigorous testing. Take, for example, work on tool use. Many researchers have considered tool technology a domain, one based in part on the functional design features of its objects. Thus, when animals such as chimpanzees and New Caledonia crows – natural tool users – show sensitivity to an object's design features, using those objects that are most likely to lead to successful outcomes, we consider this to be evidence of domain-specificity. And yet, cotton-top tamarins – a species that never uses tools in the wild and shows virtually no interest in object manipulation in captivity – show the same kind of sensitivity to an object's design features as chimpanzees and crows; furthermore, this sensitivity appears early in ontogeny in the absence of experience (Hauser et al. Reference Hauser, Pearson and Seelig2002a). This suggests that we should be more cautious with our claims of DSM capacities, and thus, how we classify the tasks within a general intelligence battery.
In conclusion, although Burkart et al. introduce a tension between DSM and DG that doesn't exist, incorrectly consider DSM perspectives as innate and inflexible, and falsely accuse other scholars of classifying nonhuman animals as rigidly DSM, they are correct in emphasizing the importance of looking more deeply at general intelligence in animals. Progress will depend on a clear articulation of the different skills tapped in the general intelligence battery, and standard methods that can be implemented across a diversity of species.
You have
Access
“The best-laid schemes o' mice an' men,” penned Robert Burns in his ode to mice. It is an astute observation of how our intelligence has outwitted theirs. Though I appreciate Burkart et al.'s attempt to synthesize a wild and wooly comparative literature on general intelligence in mice, men, and many other species, they have introduced a few red herrings and false premises that muddy the waters and undermine suggestions for future research.
The problem starts with the authors' initial premise: Scholars tend to view the mind as either domain-specific/modular (DSM) or domain-general (DG), and those who lean to DSM see the mind as predetermined and inflexible, and thus largely the work of nature. These views conflict with a theory of general intelligence. Burkart et al. claim that their framework shows “that human cognition involves elements of domain-specific and domain-general processes” (sect. 1.2.3, para. 4), and in contrast to prior views, “animal minds need not be bundles of specialized cognitive adaptations” (sect. 1.2.3, para. 4). But their premise is false as is their characterization of animal research. This commentary addresses these misconceptions and introduces some additional distinctions in order to productively explore how general intelligence evolved.
Those who have synthesized DSM perspectives (e.g., Pinker Reference Pinker1997) do not deny the existence or significance of DG capacities: evolutionarily ancient mechanisms that typically interface with and often constrain the outputs of each domain. Research on theory of mind (ToM), number, and language – domains often considered as modules – has long explored how executive functions interact with the computations and representations of each domain (Bradford et al. Reference Bradford, Jentzsch and Gomez2015; Soltész et al. Reference Soltész, Goswami, White and Szűcs2011). For example, delays in the expression of ToM and number competence are intimately related to the development of working memory, whereas performance on ToM tasks can be improved by lifting constraints that arise from inhibitory control or perseverative responses. Thus, although it is inaccurate to pigeonhole scholars as either DSM or DG, it is true that those who have explored the nature of DSM systems are more interested in them and in how they can be characterized on the basis of evolutionary theory. Similarly, although the generative computations that subserve language competence (but also other domains such as music, number, and ToM) have no limit, our capacity to produce or comprehend sentences is limited by working memory. Thus, although DSM-focused researchers tend to emphasize the nature of the representations and computations within a domain or module, they don't deny the existence or potentially constraining impact of DG processes.
Of relevance to the evolution of general intelligence is the underlying architecture of DSM systems. Here, too, Burkart et al. mischaracterize these as innate and inflexible. Research on faces reveals this error. Neurobiological studies in macaques and humans reveals dedicated circuitry that is consistent with a DSM perspective. However, this system matures slowly over time and depends on experience with faces as elegantly demonstrated by studies of individuals with early-appearing cataracts that were later removed (Rhodes et al. Reference Rhodes, Nishimura, de Heering, Jeffery and Maurer2017). A similar characterization applies to language, wherein there are core underlying computations and representations, some specific to language and others shared (Hauser & Watumull Reference Hauser and Watumull2016), but with experience selecting among the options to generate specific languages (e.g., French, English).
Lastly, it is simply not the case that nonhuman animals are perceived as mere bundles of modules, fixed and inflexible. Research on model systems such as aplysia and songbirds reveals both ancient, general mechanisms for learning and memory, as well as highly dedicated systems that nonetheless show plasticity. For example, although passerines acquire their song on the basis of specialized circuitry that enables vocal imitation, this same system requires specific input (e.g., species-specific song), is not engaged for other vocalizations (e.g., alarm calls), and in some species, shows plasticity throughout life as individuals create new songs each season. In addition, many researchers have recognized and detailed other DG processes that go beyond what Burkart et al. discuss. For example, there is considerable comparative work exploring the concept of “sameness,” analogical reasoning, and algebraic computations (Martinho & Kacelnik Reference Martinho and Kacelnik2016; Smirnova et al. Reference Smirnova, Zorina, Obozova and Wasserman2015; ten Cate Reference ten Cate2016). These are not part of the executive system, have not typically been linked to general intelligence, and yet they cut across domains and appear evolutionarily ancient.
Putting these strands together suggests that any approach to exploring the evolution of intelligence must consider the interaction between DSM and DG, understand the specificity of the content of DSM, examine a diversity of DG systems (i.e., beyond executive functions), and document how maturational changes in DG can impact the ontogeny of DSM. The content of a domain is particularly relevant as tasks within the general intelligence battery are often assumed to be part of a given domain without rigorous testing. Take, for example, work on tool use. Many researchers have considered tool technology a domain, one based in part on the functional design features of its objects. Thus, when animals such as chimpanzees and New Caledonia crows – natural tool users – show sensitivity to an object's design features, using those objects that are most likely to lead to successful outcomes, we consider this to be evidence of domain-specificity. And yet, cotton-top tamarins – a species that never uses tools in the wild and shows virtually no interest in object manipulation in captivity – show the same kind of sensitivity to an object's design features as chimpanzees and crows; furthermore, this sensitivity appears early in ontogeny in the absence of experience (Hauser et al. Reference Hauser, Pearson and Seelig2002a). This suggests that we should be more cautious with our claims of DSM capacities, and thus, how we classify the tasks within a general intelligence battery.
In conclusion, although Burkart et al. introduce a tension between DSM and DG that doesn't exist, incorrectly consider DSM perspectives as innate and inflexible, and falsely accuse other scholars of classifying nonhuman animals as rigidly DSM, they are correct in emphasizing the importance of looking more deeply at general intelligence in animals. Progress will depend on a clear articulation of the different skills tapped in the general intelligence battery, and standard methods that can be implemented across a diversity of species.